Modifying QPPM to measure "Max. True Peak Level"

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

electro_aLex

Member
Joined
Aug 15, 2011
Messages
14
Hello together,

i had I discussion with a good friend about some analog-audio theory, especially our discussion was about PeakProgramMeters.

My thoughts were: the old peak meters we know, measure the "real peaks" (full-wave-peak-rectifyeng) and multiply it with 0,71 (RMS-value).
My friends opinion: the signal runs through a full-wave-peak-rectifier, which simultaneously generates the RMS-value.

We are agreed that:
This works good for perfect sine-signals, but not for music- or human language -signals (problems with short and high signals).
Which opinion is correct?

With the new EBU Recommendation R128, witch is the new standard in television broadcasting since 31.08.2012
these old Quasi Peakmeters are worn-out.

The old QPPM-values are load-related - what is the background in relation to this fact?
Are there any advantages in measuring this way?

My question is: is it possible to bypass the RMS-converter of the old QPPM's to display the "true" peak, like the new EBU-standart recommends?
Or is there just a (adjustable) gain adjusted to x0,71 that can be set to x1?


Is there someone who can explain exactly the processes inside a QPPM (like the often used NTP/RTW plasma QPPM's)?
I think I misunderstand a lot of this stuff, for example:
I saw a camparison grapic between QPPM and TPPM, the 1Khz sine test tone gives exactly the same level on both meters - but why?
Isn't the QPPM-value just 71% instead of 100%?


Best regards,
Alex
 
The difference of "quasi" and "true" is a matter of attack time. A "quasi" takes a few ms to follow a fast peak, a "true" follows as fast as it gets. The latter gives a better indication of overload (important for ADCs/DACs), the former is perhaps visually a bit more pleasing.

Samuel
 
electro_aLex said:
Hello together,

i had I discussion with a good friend about some analog-audio theory, especially our discussion was about PeakProgramMeters.
I have been messing with combination peak/ave (VU) meters for decades.
My thoughts were: the old peak meters we know, measure the "real peaks" (full-wave-peak-rectifyeng) and multiply it with 0,71 (RMS-value).
There are sundry legacy peak meters ranging from gas (neon?) lamps that can be arbitrarily fast, to PPM with an intentionally slowed response since real time was perceived as faster than useful (for broadcast and media of the time).

How the peak output is scaled is arbitrary and can be a complex topic depending how you want to parse it. Since I displayed both peak and average on the same display it was useful for me to present them with the typical 3dB spread.
My friends opinion: the signal runs through a full-wave-peak-rectifier, which simultaneously generates the RMS-value.
? Peak detection and RMS weighting/conversion is an independent operation, but this may be getting into semantics.
We are agreed that:
This works good for perfect sine-signals, but not for music- or human language -signals (problems with short and high signals).
Which opinion is correct?
The math to use sine-wave equivalent weighting for more percussive transient program material is just an arbitrary decision that seems as good as any other.

What are you trying to accomplish from metering? We now have enough cheap computer power that we might be able to apply complex weighting functions to better represent perceived loudness, but most metering has simpler goals, like keeping signals within the dynamic range of the media involved. One could argue that PPM with some 4mSec attack time could be faster for digital paths, while 4 mSec of clipping could equally be argued as not audibly significant (i would prefer fast for digital recording). 
With the new EBU Recommendation R128, witch is the new standard in television broadcasting since 31.08.2012
these old Quasi Peakmeters are worn-out.
R128 looks like it is addressing perceived loudness, not signal headroom wrt peak levels within hardware. This looks like an attempt to give a more robust metering standard to use to help normalize signals to sound similar for loudness. 
The old QPPM-values are load-related - what is the background in relation to this fact?
Are there any advantages in measuring this way?
I do not understand "load related"?
My question is: is it possible to bypass the RMS-converter of the old QPPM's to display the "true" peak, like the new EBU-standart recommends?
Or is there just a (adjustable) gain adjusted to x0,71 that can be set to x1?
I don 't know that peak meters need to be tweaked, this issue seem to be about variance between peak levels and perceived loudness.

This is not a new concept for me as I have been supplying peak and VU (or ever looked at RMS) simultaneously. While even this is not a perfect loudness meter. Dolby has done a bunch of work to help normalize loudness after the fact for consumers to help avoid the annoying commercial blare.

I am not sure it possible to come up with a completely objective loudness meter since human perception is involved. Loudness for tones is well studied, but loudness for complex sounds is well, complex. 
Is there someone who can explain exactly the processes inside a QPPM (like the often used NTP/RTW plasma QPPM's)?
I think I misunderstand a lot of this stuff, for example:
I saw a camparison grapic between QPPM and TPPM, the 1Khz sine test tone gives exactly the same level on both meters - but why?
Isn't the QPPM-value just 71% instead of 100%?


Best regards,
Alex
I am not familiar with that nomenclature, and too lazy to google to find out. Most likely differences are attack times. Not scaling.

It sounds to me like this is not about peak levels but loudness which is a different animal.

JR
 
Samuel Groner said:
The difference of "quasi" and "true" is a matter of attack time. A "quasi" takes a few ms to follow a fast peak, a "true" follows as fast as it gets. The latter gives a better indication of overload (important for ADCs/DACs), the former is perhaps visually a bit more pleasing.

Samuel

No, I dont think so. A lot of "quasi's" allow to measure in an integration-time of 0,1/0ms ("fast" or "zero"-mode on some NTP 477s).
What is the real difference?

I found a .pdf in relation to these questions: https://www.tcelectronic.com/media/LundT013011.pdf
 
"Quasi-peak Meter Display of audio level. The quasi peak meter, PPM or QPPM, has been widely used in broadcast
and film. Standardized as IEC 60268-10, the meter is intentionally slowed down in its “attack”
response time, and even more in its decay. Arguably, it’s neither a peak level nor an average level
meter, and also not suitable for measuring loudness level. Higher peaks than what is shown on a
QPPM can readily be found in a signal when also measured by a true-peak meter."

"True-peak Level An absolute measure used to describe the true-peak level of a digital signal. Can be measured as
specified in ITU-R BS.1770. The intrinsic or true-peak level of a digital signal may be 3 or more
dB higher than the sample peak level of the same signal. Contrary to sample-peak level, true-peak
is a valuable estimate of the headroom required to handle a signal without clipping. 0 dBFS+ level
is routinely printed to CDs, leading to distortion in reproduction equipment, sample rate converters
and data reduction systems. For more information see Lund, “Stop Counting Samples”, AES 121
convention paper."

Furthermore I found the "RECOMMENDATION  ITU-R  BS.1770
Algorithms to measure audio programme
loudness and true-peak audio level"

"Annex 2

Guidelines for accurate measurement of “true-peak” level
This Annex describes an algorithm for estimation of true-peak level within a single channel linear PCM digital audio signal. The discussion that follows presumes a 48 kHz sample rate. True-peak level is the maximum (positive or negative) value of the signal waveform in the continuous time domain; this value may be higher than the largest sample value in the 48 kHz time-sampled domain. The algorithm provides an estimate for the signal as it is, and, optionally, as it would be in the event that some downstream equipment were to remove the DC component of the signal. Optional mild high frequency pre-emphasis in the peak measurement signal path can enable the algorithm to report a higher peak level for high-frequency signals than is actually the case. The purpose for this is that the phase shifts of subsequent signal processing stages (such as Nyquist filters) could cause growth of high frequency signal peaks, and in some applications this feature could be useful to provide further protection from downstream clipping."

This recommendation is just for digital signals, I know.

I thought here are a lot of electronic-prodigies hanging arround?
What can they advise me to do with my analog-QPPM to measure the true peaks? Is it a big problem?! :)

Thanks for the replies and best regards,
Alex
 
> Is it a big problem?

No.

Quasi-peak is valid because the human ear does not hear short distorted peaks. Various studies show that 1mS of clipping is un-heard, 5mS clipping is usually not noticed, and longer periods of clipping may not be annoying.

ALL-peak was necessary in a few situations. RCA optical film sound would "clang" if the light valve went to 100% shut, even for an instant. RCA developed a neon-lamp "meter" so all clanging could be avoided.

Most audio systems have high headroom or benign clipping. In AM radio, high average level is needed to overcome atmospheric noise, but the 25,000 Watt modulator "can" be over-driven for 1mS or 10mS and listeners won't notice.

Now with generally digital audio of sufficient bit-depth, we do not have to slam the system to stay away from noise. Rather than pushing average up to 10dB below clipping and tolerating clipping 1% of the time, we can set average level 16dB, 18dB, even 20dB below clip (digital maximum) and still keep noise very low.
 
I do not understand "load related"?

With this expression I mean, that the displayed value with a QPPM is not just a voltage (like in a true peak meter), it is the "average electrical power" (just the x0,71-value of the real peak, related to the integration time).

My question was not about WHY true peak-measuring is necessesary or not, it was about HOW changing a common PPM to measure true peak.

I found some schematic documents about the NTP 277-400, and how different stages are connected with each other.
The last stage, called "level converter" / "output amplifier" seems to me like the RMS-converter, am I right with this?

If we bridge the level converter we should get the true peak (in combination with a "zero" integration time)?
Or if we are able to adjust the gain of "output amplifier" to x1 (trimpotentiometers in the 177-400 > "ref. level"), we could also get the true peak value.

I found a document of a previous NTP Peakmeter model 177-400 on the net, too.
http://uebergabe-daten.hifi-classic.at/httpdocs/BDA_SM_Archiv/NTP/NTP_177-400_$$$_BDA-SM.pdf

Maybe this helps to identify the processes inside a QPPM.

Best regards,
Alex
 

Attachments

  • NTP 277-400 manual.jpg
    NTP 277-400 manual.jpg
    245.4 KB · Views: 30
In the middle of the left block diagram there is section labelled "integration time circuit", this would be applying a few mSec of lag to meet PPM standard.

JR
 
The integreation time isnt the problem at all.

I need to know exactly what part of the circuit I have to bypass to get the true peak level, and this cant be wizardry ;-)
 
Once again we appear to have a failure to communicate.

To realize a true =fast peak, you need to reduce or eliminate the integration time constant (roughly 4 mSec in PPM).

Asked and answered.  Why do I think repeating this will not help?

JR
 
NO! Here the translated version of the German Wikipedia artice "Aussteuerungsmesser" (Peak Program Meter):

RMS Level Meter [Edit]

An RMS power meter integrates the measured values ​​over the design speed measurement period, with standardized quasi-peak power meters, these are 5 or 10 ms.

With a standardized quasi-peak level meter of the reference point ("0 dB") corresponding to the rms value of a sinusoidal signal (· USS; corresponding ≈ 0.707 · USS) scales, ie by 3 dB. In the case of a conventional Messtons (continuous sinusoidal signal) is therefore of the level value of the quasi-peak voltage must numerically not be distinguished from that of the peak voltage.

Music, speech or signals such as rectangular or sawtooth arising measured values ​​deviate from the display of a pure peak level meter. For one-time, very short duration signals are obtained by the dynamic properties of the display significantly lower readings. A continuous wave signal is displayed on the other hand with a quasi-peak level meter by 3 dB louder than with a peak level meter (True Peak Meter). This property leads to the display at the first glance curious displaying +3 dBFS for a full-level square-wave signal. The RMS power meter assumes the transmission range no frequency weighting.
 
electro_aLex said:
Hello together,

i had I discussion with a good friend about some analog-audio theory, especially our discussion was about PeakProgramMeters.

My thoughts were: the old peak meters we know, measure the "real peaks" (full-wave-peak-rectifyeng) and multiply it with 0,71 (RMS-value).
My friends opinion: the signal runs through a full-wave-peak-rectifier, which simultaneously generates the RMS-value.

Alex

I think neither is correct.

Most 'old' PPMs used a full wave peak rectifier circuit and that's it. The only 'multiplication' would be a gain calibration used to set the meter to some specific level using a sine wave input. The attached screenshot show how the PPM meter drive circuit was done in an old Studer mixer. The input signal and its inverse are peak rectified and applied to a buffer to driver the meter movement. Circuit values set the attack and decay times. Provided the attack time is shorter than one cycle of the highest frequency (20KHz) and the decay time is longer than the lowest frequency (20Hz) the peak detector will show the true peak of the input signal. In practice the attack time constant is set to hold the peak value long enough for the meter movement ballistics to allow the pointer to actually reach the indicated value (usually 4 to 10mS) and the decay time is set to somewhere between one and two seconds.

Cheers

Ian
 

Attachments

  • Screenshot from 2012-11-18 17:22:39.png
    Screenshot from 2012-11-18 17:22:39.png
    144.6 KB · Views: 28
electro_aLex said:
My thoughts were: the old peak meters we know, measure the "real peaks" (full-wave-peak-rectifyeng) and multiply it with 0,71 (RMS-value).
My friends opinion: the signal runs through a full-wave-peak-rectifier, which simultaneously generates the RMS-value.
As a complement to Ian's answer, both your opinions are incorrect because:
a) "multiply it with 0,71" is meaningless; you have to take into account the sensitivity of the meter itself (be it electromagnetic or bargraph-based). You don't have to actually multiply; the 0.707 factor will be embedded in the final transduction factor.
b) There's no way you can actually extract rms value from peak-value. RMS computing needs to know the complete history of the signal, if you apply a 1st-order hold (as in peak detection), you lose all history.
The 0.707 factor has been introduced in order to offer user's compatibility with the historic mean-value VU-meter - remember it is itself compensated for rms compatibility by a factor 0.9 (mean value of a sinewave is 2.sqrt2. Vrms/pi).
In jurassic times, the VU-meter was the only convenient meter; it is well known that the ballistics of the VU-meter are useless for controlling accurately signals approaching the threshold of clipping of most electronic equipment, but with some ingenuity, operators had developped a mastery of level control based on their aural evaluation of the signal's peak-factor. Magnetic tape was particularly tolerant to overload, so VU-meters were the norm, but for disc and film mastering, QPPM's were mandatory. Even then, it was necessary to add protection against ultra-fast transients, in the form of clippers.
The EBU recommandation addresses a different subject, the so-called "intersample overs", which happen basically because of the transient response of low pass filters, in particular the anti-aliasing filters in converters and the band limiting filters in the broadcast transmission chain.
 
> The 0.707 factor has been introduced in order to offer user's compatibility with the historic mean-value VU-meter

Yes; and also for checking against a workbench meter (such as a Fluke) which is traditionally intended for Power Line measurements and assumes Sine and assumes you want to know "RMS". (Some are internally Average and calibrated 1.1; others are Peak and calibrated 0.7.) When my board meter is reading +4dBu, I want my Fluke to say 1.228V. True *only* for a good Sine, of course. Reading live audio with bench-meters is pointless.
 
abbey road d enfer said:
a) "multiply it with 0,71" is meaningless; you have to take into account the sensitivity of the meter itself (be it electromagnetic or bargraph-based). You don't have to actually multiply; the 0.707 factor will be embedded in the final transduction factor.
b) There's no way you can actually extract rms value from peak-value. RMS computing needs to know the complete history of the signal, if you apply a 1st-order hold (as in peak detection), you lose all history.

Ok, you confirm the fact that the values we get displayed on the QPPMs are influenced because of the 0,707 factor which is embedded in the final transduction factor - and the fact that these values are "wrong".
And my whole desire is, to convert a QPPM into a TruePPM :)

HOW can we eliminate this 0,07 factor?  There should be a single part of the circuitry that is responsible for the special transduction factor.

Thanks for the answers!

Brgds,
Alex
 
electro_aLex said:
Ok, you confirm the fact that the values we get displayed on the QPPMs are influenced because of the 0,707 factor which is embedded in the final transduction factor - and the fact that these values are "wrong".
I didn't say they're actually "wrong", but I understand that it's not what you want to see. As JR suggested, you may work on the basis that you consider -3fs (whatever it is on the scale) as your absolute maximum. Measuring "absolute" peak is useful only if you know what your maximum is and if you see this maximum (more or less at the top of the scale); measuring absolute peaks with a scale that goes to +10 is not very relevant (unless you have a system where operating level is only 10dB below absolute max). In that respect, the DIN scale (-50 to +5) is somewhat meaningless, the Nordic scale (-60 to +9) being just a little more satisfactory. IMO, the only valid scale for a peak meter is the FS scale. Whatever the ticks, you want as much resolution as can be in the -30 to 0dBfs range.
And my whole desire is, to convert a QPPM into a TruePPM :)

HOW can we eliminate this 0,07 factor?  There should be a single part of the circuitry that is responsible for the special transduction factor.
In RTW meters there is a trimmer that allows adjustment of sensitivity. It should provide enough range for achievng what you want. I suspect NTP's offer the same possibility. The owner's manual is your friend, as ever.
 
I think it is important to state the reason for the metering in question.

If it is in order to meet some broadcast standard, or measurement for some purpose legal or regulatory... well... that's no fun, and there is no reason that the law or standard has to make sense.  So if that is the purpose, then it makes sense why this thread seems to go round and round.

On the other hand, If it is to measure an actual audio signal and use that measurement to manage levels in real time, then all useful meters have a decay time for display, and ideally a faster attack time.  And a very fast clip indicator, which stays lit is also nice.

I have used meters primarily for setting mic gain and what I like is:
The fast attack is needed to observer the levels the transients are washing over.
The slow decay is so that your eyes can have time to see what the fast attack is measuring.
The clip indicator is a warning light ideally set to a level where you think overload is likely or imminent.

To do this, VU is too slow, VU and a clip light is too slow, Mechanical meters are pretty, but too slow or too expensive (pick one).  LED is cheap, visible and works fine from across a room.

I have never tried to make a meter using a computer, which would give all sorts of additional options, but for real time use I would think it would still have an attack and decay time... I suppose it could keep and display a record of other transients, including what is being called "true" peak (which I can only assume means the peak instantaneous positive or negative voltage of the signal).  Although I am not sure if that true peak would really be a useful measurement.

The notion of a peak level seems to be misunderstood.  Our signals are AC, and the positive or negative peak of those AC values is both very hard to measure using analog only (digital could do it), and is pretty useless from a practical point of view.


 
bruce0 said:
I think it is important to state the reason for the metering in question.
You're absolutely right. I assumed from the OP that he was concerned about preventing clipping. Maybe I'm wrong...
Although I am not sure if that true peak would really be a useful measurement.
The notion of a peak level seems to be misunderstood.  Our signals are AC, and the positive or negative peak of those AC values is both very hard to measure using analog only (digital could do it), and is pretty useless from a practical point of view.
Level meters are not measurement tools in the metrological sense. They are not used to quantify and record a value, they are tools for the evaluation of a physical quantity, in the course of a process where exceeding a value may be detrimental. Just like the speedo in a car is not a test instrument, it's an indicator that is used in the feedback process that includes the driver's foot, gas pedal, fuel injection...for the regulation of speed. As such, it does not belong to the same category as measurement instruments, its output needs to be presented to the operator in a way that's simplified and meaningful. A level meter is much more efficient in the form of a multicolor bargraph than a numeric display, although the latter can be much more accurate.
Measuring peaks in analog is not hard, what is somewhat harder is presenting the resulting information in a convenient way, and depends on the application.
 
Enfer:

I agree that Meters are indicators not measurers. ( Except if you are trying to comply with a regulation, or make your advertisements louder on TV! )

Form follows function, metering you might want for a front of house board is very different than what you might want on mic channel inputs or during mixing.

And agreed it is not "hard" to measure, what I meant is that the typical analog circuits that rectify and "store" peaks use opamps, and use capacitors which take some time to charge, and awkward things start to happen at high frequencies. A more complex circuit can accomplish it pretty well at audio frequencies but there are tradeoffs.  If you really want to measure and store peak, better do it in the digital domain, and get all the advantages of display options.  But as I said and I think you agree, knowing the level of a true instantaneous peak, is not very useful.

What I normally want out of my meters is to know whether a singer has been pushing my levels too far when I glance away, and whether the shy one has stepped back from the mic again, and whether I have too much guitar getting into my vocal mics, etc.  The meters are painting a picture that is fast enough to be a sharp picture, but slow enough for me to see it.  The picture is of what is going on in the sound field and what can I do to better capture the music.

bb
 
I am still at a loss over what the OP is trying or expecting to achieve?

I have been designing meters for a few decades, but won't further clutter this thread with my blather.

+1 to most points already made, but still "what's the frequency Kenneth"?

JR
 
Back
Top